CN102074240B - Digital audio watermarking algorithm for copyright management - Google Patents

Abstract

The invention relates to a method for performing copyright management by utilizing digital audio watermarking. The method comprises the following steps: embedding a watermarking signal in a logarithm domain of the audio signal energy so that the decoding is unrelated to the amplitude; reasonably designing the frame structure of the embedded watermarking information so that the timing statistical accuracy of a de-embedded result reaches second, and the decoding time accuracy and decoding accuracy rate are not influenced under the operations of cutting, splicing and inserting other audio signals and the like; encrypting the embedded information by an RSA (Asymmetric Algorithm) digital signature algorithm so that the embedded information has security; and effectively reducing interference of the carrier audio to the watermarking information by the embedding algorithm by using the relativity of the short-time stability of the audio so as to obviously improve the success rate and accuracy rate solved by the watermarking. In the method, a digital watermarking technology is introduced to identify the audio materials in a medium library, and then the broadcast flow or audio file are decoded and analyzed to obtain the service conditions of materials, thereby acquiring copyright payment information.

Description

A kind of digital audio watermarking algorithm that is used for copyright management

Technical field

The present invention proposes the Audio Watermarking Algorithm that a kind of matchmaker of being used for supports copyright management.

Background technology

The digital audio frequency watermark technology is originated, is carried out the entitlement authentication and follow the trail of unauthorized behavior being widely used in the authentication multimedia as an important directions of field of information security technology, has very wide application prospect ^[1]In the middle of the program making process of radio and television etc., hope can be carried out statistical study timely to the operating position of multimedia material, so that carry out the management that media article and copyright are used.Through in multimedia document, adding the watermark information that is used to identify specific material, obtain the operating position of materials related afterwards through watermark decode, can more convenient, strong support be provided for the matchmaker supports management process.

Based on the needs of reality, we need watermark to support copyright management system the matchmaker has following characteristics.1) can identify copyright, matchmaker's standing breath, the watermark information that needs to embed has certain information capacity.2) have certain security, avoid unauthorized side to recover and the modification watermark.3) stronger robustness.In the use of audio content, may be some editors to audio frequency, like the shearing of optional position, splicing, MPEG compression or the like.Require the higher robustness of watermark of embedding, passing through editing operation and can tell under the situation of original contents, also can solve embedding information smoothly and can obtain duration information thus.4) can confirm the duration of embedding information, and have the regular hour precision.So that the content of playing is carried out the duration statistics.Program management service time that is used for needs are paid etc.

Proposed from people such as Tirkel since the notion of digital watermarking, in short time more than ten years, the development of digital watermark technology can be described as with rapid changepl. never-ending changes and improvements, and progressively deep and ripe.The development of digital watermark technology comprises test and the judge of design, digital watermarking attack, the digital watermarking algorithm of digital watermarking algorithm, theoretical research of digital watermark technology or the like.The algorithm of the spread-spectrum watermark that people such as Cox propose has milestone inthe in the digital watermark technology development.People such as Cox think that digital watermarking not only will reach not sentience, promptly realize covert communications, also will resist various attack, have robustness.Therefore he thinks, when guaranteeing not sentience, be embedded into watermark information the main part of perception of host signal.Afterwards, various robustness algorithms emerge like the mushrooms after rain.At present, on the modulation system of watermark, generally be divided into watermarking algorithm the quantization index modulation watermarking algorithm (QIM) of people's propositions such as the spread-spectrum watermark class algorithm that proposes based on Cox and Chen.Because spread-spectrum watermark carrier audio frequency when decoding can have certain influence to watermark signal, thereby therefore when decoding, need manage to reduce the robustness that this influence improves watermark.Because the influence of carrier audio frequency, the code check of the embedding of spread-spectrum watermark method receive some restrictions, the time precision when through watermark material being used statistics is not high.In addition; Can cause the watermark information of whole one-period to be lost if in a single day error code appears in the synchronizing signal of watermark; Make the appearance a lot of watermark informations of watermarking algorithm after common multi-media edit operation such as random shearing, splicing, insertion therefore lose, thereby further have influence on duration accurate rate material operating position statistics.

Summary of the invention

The objective of the invention is to; Exist the robustness of embed watermark information and time precision not high for overcoming present algorithm; Watermark information is lost seriously in operations such as the random shearing that carries out multi-media edit, splicing back; The operating position that causes very difficult watermark information analysis to embedding to obtain the multimedia material is carried out problems such as precise statistics, and the present invention proposes the Audio Watermarking Algorithm that a kind of matchmaker of being used for supports copyright management.

For achieving the above object, the present invention proposes a kind of method of carrying out copyright management with digital audio frequency watermark, this method comprises the extraction detection method of the embedding grammar and the watermark information of watermark information, wherein,

The embedding grammar of described watermark information comprises following steps:

(1-1) watermark information pre-service is at first encrypted watermark information and then is encoded, and obtains watermark encoder information sequence { e _i, 0≤i＜L _B, L wherein _BLength for the watermark information coding;

(1-2) watermark encoder information { e to obtaining _i, 0≤i＜L _BCarry out framing encapsulation;

At first, the watermark encoder information that the matchmaker supports copyright management information that contains that embeds is divided into L _sSection, and with the different different segmentations of spreading code sign; Then, all add a bit synchronization position before the segmentation with each watermark encoder information, information bit adopts different spreading codes with synchronization bit, obtains watermark and embeds information { w _i, 0≤i＜L _S+ L _B;

(1-3) watermark that obtains is embedded information { w _i, 0≤i＜L _S+ L _BSequence carries out base band balance modulation, the watermark with a bit during modulation embeds Information hiding among the sound signal of continuous several frames, obtains the signal S after the balance baseband modulation;

(1-4) signal S is modulated in the original audio signal, the modulation formula is following;

Y(n)＝X(n)·10 ^{w(n)·s·α(n)}

Wherein, { X (n), 0＜n＜N} treat the amplitude of the sound signal of embed watermark at frequency domain, and N is the length of the spreading code of use, and X (n) unit is dB; { w (n), 0＜n＜N} are spreading code, and this spreading code is that information bit or synchronization bit are confirmed by the type that embeds bit, and α (n) calculates through psychoacoustic model, are used for control and embed energy;

(1-5) combine original phase information to do the IFFT conversion again the signal Y (n) of embed watermark at last; Accomplish the embedding of a frame, will accomplish again between each frame that watermark embeds and do overlap-add, recover time-domain signal; Obtain the sound signal of embed watermark, thereby accomplish the telescopiny of watermark;

The extraction and the detection method of described watermark information comprise following steps:

(2-1) relevant decoding, the process of this process and coding is opposite, obtains false-alarm probability and false dismissal probability through relevant decoding;

(2-2) synchronizing process, this synchronizing process comprises frame synchronization and data sync, is used to solve the information code of synchronous segmenting;

(2-3) segment information splicing, the segmentation that will occur and meet some completion decodings of periodic regularity according to the splicing decision condition continuously is stitched together, and obtains information service time of audio frequency;

Wherein, the splicing decision condition is:

1) frame synchronization position Pos ₂Frame synchronization position Pos with a last segmentation ₁Meet | Pos ₂-Pos ₁-L _sL _f|＜L _Sync, L _fBe the length on the segments theory, L _SyncConfirm according to experiment;

2) the spreading code sequence number that obtains for before the spreading code sequence number next one sequence number that should occur;

3) decoded result coming to the same thing of previous periodic segment with it;

Satisfy above three conditions simultaneously and just splicing implemented in segmentation, and calculate splicing first be fragmented into the sampled point distance of last segmentation, obtain the service time of audio frequency then through sampling rate;

(2-4) watermark information reduction, the extraction of accomplishing watermark information detects.

As an improvement of the present invention, described synchronizing process adopts correlation properties to reduce mistake synchronously, and concrete steps are:

Limit score S _wSurpass a certain threshold value, this CALCULATION OF PARAMETERS formula is following:

$S_w=\underset{k}{\mathrm{\Σ}}C\left(k\right)\·W\left(k\right)$

Wherein, C (k) is the decoding score;

W (k) is a window function, and computing formula is following:

$W\left(k\right)=\left\{\begin{array}{c}1,\left(\right|k-k_{\mathrm{sync}}|\&le;L_p/2)\\ -1,(L_p/2<|k-k_{\mathrm{sync}}|\&le;L_p)\end{array}\right.$

Wherein, L _pThe mean distance at the forward peak that obtains for statistics and reverse peak, k _SyncBe the sync bit point.

Technique scheme, described watermark information pre-service concrete steps are:

Rule below at first utilizing produces the PKI e and the key d of RSA Algorithm, and specifically rule is as follows:

Choose at random two big prime number p and q, p is not equal to q, calculates M=p * q; According to Euler's function, be not more than M and be (p-1) * (q-1) with the relatively prime integer number of M; Select an integer e with (p-1) * (q-1) relatively prime, and e is less than (p-1) * (q-1); Calculate d with following this formula:, d * e ≡ 1 (mod [(p-1) * (q-1)]) destroys the record of p and q;

The encryption formula that utilizes RSA Algorithm then is to the watermark information sequential encryption;

Information { c after the encryption _iThrough Bose-Chaudhuri-Hocquenghem Code, form the watermark encoder information { e that encrypts _i, 0≤i＜L _B.

Technique scheme, described synchronizing process adopt the relevant method of sliding in the process of search synchronization bit, step is:

At first, setting a sliding step is correlated with; The position of maximum of then, getting relevant score is as the position of frame synchronization; At last, find suitable frame lock bit postpone, carry out data sync, the reference position of specified data.

Technique scheme, described watermark information reduction comprises following steps:

The decoded result sequence of at first, establishing the segmentation that stitchs together is { e _i', earlier this result is carried out the BCH decoding and obtain information { c _i'; Then, again with watermark information { c _i' decipher through RSA Algorithm, obtain m '; At last with m be expressed as binary sequence m ' _i, the decrypting process of completion information.

Technique scheme, described RSA Algorithm decryption step is:

Earlier with { c _i' be expressed as integer c ';

Then, decipher based on the deciphering formula of RSA Algorithm, it is following specifically to decipher formula:

c′ ^d＝m′(modM)，

Wherein d is a key, and preceding with it definition is identical.

Technique scheme, described spreading code are balance Gold code.

Technique scheme when described segment information splices for some isolated concrete treatment step of segmentation that occurs is:

, segmentation is spliced to together when meeting following condition simultaneously;

1) frame synchronization position Pos ₂Frame synchronization position Pos with a last segmentation ₁Meet | Pos ₂-Pos ₁-L _sL _f|＜L _Sync, L _f, L _sWith L _SyncDefine identical with before;

2) the spreading code sequence number that obtains for before L after the spreading code sequence number _sThe individual sequence number that should occur;

3) decoded result coming to the same thing of previous periodic segment with it.

The invention has the advantages that this algorithm utilizes the psychologic acoustics masking effect, make the watermark that embeds adjust, to guarantee the not sentience of embedding information according to acoustic characteristic.Algorithm utilizes spread spectrum ^[2]The relevant in short-term characteristic of method and sound signal improve the robustness of embed watermark.And to the copyright application, through the unique matchmaker of material embedding is supported copyright information material is identified, obtain the operating position of multimedia material then through watermark decode.Algorithm reaches very high time precision through reasonable arrangement information frame structure and adjustment synchronization policy.Through the sense of hearing test shows of formal strictness, our algorithm has very high not sentience.The robustness of carrying out simultaneously test verified again watermarking algorithm can effectively resist immediately shear, the common operation in the audio frequency manufacturing process such as splicing, MPEG compression.

Description of drawings

Fig. 1 is the information frame structural drawing that provides of the present invention;

Fig. 2 is the synoptic diagram of the balance modulated process of the present invention's employing;

Fig. 3 is the process flow diagram of the watermark embed process of employing of the present invention;

Fig. 4 is the process flow diagram of the method for synchronization of the present invention;

Fig. 5 is the peak value simulation drawing that relevant score of the present invention occurs.

Embodiment

Below in conjunction with accompanying drawing and embodiment the present invention is further specified.

This paper has proposed to be used for the method for the audio frequency watermark of media article and copyright management.This algorithm has following characteristics: 1) utilize the principle embed watermark information of spread spectrum, improve the anti-interference of embedding information.2) utilize the psychoacoustic model of MPEG to guarantee the not sentience of the information of embedding.3) information that embeds is carried out the encryption and the randomization of the information of carrying out with RSA Algorithm, and enciphered message is carried out the BCH Error Correction of Coding, further improve the security of information.4) adopt the structure of balance modulation, utilize the steady characteristic in short-term of sound signal, reduce of the influence of carrier audio frequency watermark.When 5) embedding watermark information is embedded continuously, obtain the service time of material then according to decoded result.6) for improving the temporal resolution of decoded result, with the watermark information segmentation that embeds, and with the different different segmentations of spreading code sign.During decoding, the segmentation that the context relation of utilization decoding and certain decision rule will belong to same information stitchs together, and obtains the service time of material according to this.Like this, when a certain segment information is lost, can not have influence on the decoding of whole complete information section, improve the time precision of decoded result duration statistics, can make the time statistical precision of watermarking algorithm reach a second level.7) when organizing the watermark information frame, adopt a bit synchronization position afterwards with the structure of some information bits, information bit adopts different spreading codes with synchronization bit.Like this, when doing sync correlation, utilize the peak structure that produces when being correlated with; Make sync bit confirm more accurate, in addition, with this structure also as synchronous characteristic; When having only this characteristic apparent in view; Just be judged to be and correct effectively reduced mistake synchronously synchronously, and in extracting the watermark process, do not need original audio frequency and watermark information.

1, watermarking algorithm introduction

1. watermark information pre-service

At first, suppose that the watermark information sequence is { m _i, 0≤i＜L _m, L _mBe the watermark information sequence length, this information is generally copyright information, as the unique identification of waiting to embed material file.Produce RSA Algorithm with following rule ^[3]PKI e and key d:

(1) choose at random two big prime number p and q, p is not equal to q, calculates M=p * q.

(2), be not more than M and be (p-1) * (q-1) with the relatively prime integer number of M according to Euler's function.

(3) select an integer e with (p-1) * (q-1) relatively prime, and e is less than (p-1) * (q-1).

(4) calculate d:d * e ≡ 1 (mod [(p-1) * (q-1)]) with following this formula.

(5) record of p and q is destroyed.

Subsequently, with watermark information { m _iTransfer an integer m to less than M, so, when selecting the RSA Algorithm parameter, make that the number of bits size of M should be greater than the watermark information code length, establishing the M number of bits is L _M, L then _m＜L _MThen, according to the encryption formula of RSA Algorithm:

m ^e＝c(modM)

(1)

Then c is expressed as binary sequence { c _i, 0≤i≤L _M.

With the information { c after encrypting _iThrough Bose-Chaudhuri-Hocquenghem Code ^[4]After, form watermark encoder information { e _i, 0≤i＜L _B, L _BBe Bose-Chaudhuri-Hocquenghem Code length.Parameters such as the code length of Bose-Chaudhuri-Hocquenghem Code, error correction figure place are selected according to reasonable application.Watermark encoder information { e _iTogether with synchronizing information { s _i, 0≤i＜L _SLump together, form final watermark and embed information { w _i, 0≤i＜L _S+ L _B.

2. watermark information frame institutional framework

If it is L that watermark embeds the figure place of information, and complete representation length to be the watermark information of L need duration is the carrier audio frequency of s, as s during greater than 1 second, if during this information corresponding synchronous information dropout, whole segment information also can be lost.Like this, when adding up duration according to the watermark decode result, statistics can only be the integral multiple of s.For improving the time precision of statistics, watermark information is divided into L _sSection, every segment information length is F=L/L _sSelect L _sMake the carrier audio frequency length of representing this length watermark information less than 1s, like this, when this fragment loss,, can improve the temporal resolution of duration statistics thus if make solving of other segmentations not influenced by it through certain decoding policy.Decoding policy after be described in detail.

Before every segment information position, insert one as synchronizing information, synchronization bit adopts different spreading codes to embed with information bit, simultaneously, between the synchronization bit of different segmentations, also representes with different spreading codes.During decoding, can confirm which section this segmentation is positioned in information code through the sequence number of spreading code.

The information frame structure that draws thus is as shown in Figure 1.

3. balance modulation

Because sound signal generally has in short-term characteristic stably, so the amplitude of the frequency domain between adjacent several frames is relatively stable.Can utilize this characteristic to reduce of the influence of carrier audio frequency to watermark.During embedding, with the Information hiding of a bit among continuous several frame signals.When the bit that embeds was 1, the baseband signal S of the signal that preceding K frame is to be embedded was made as+1, and the baseband signal of ensuing K frame is made as-1, when embedding bit when being 0, the value reversed in order of S.Characteristic stably in short-term during decoding, the signal of preceding K frame deducted the signal of back K frame on log-domain, owing to can weaken the influence of carrier signal to watermark to a certain extent.

The modulation of Fig. 2 balance

4. utilize the spreading code telescopiny

When sound signal is handled, the signal of input is done the branch frame and handle, establishing the branch frame length is Fr.Do the FFT conversion after this frame added Hanning window, the top n coefficient of conversion is used for watermark and embeds, and N equals the length of spreading code.If treating the amplitude of the frequency-region signal of embed watermark is that { X (n), 0＜n＜N}, spreading code are { w (n), 0＜n＜N}.Transfer X (n) to dB value x (n)=log (X (n)), then to passing through formula (2) embed watermark.

y(n)＝x(n)+w(n)·S·α(n)

(2)

Wherein, S is the base band coded signal of this frame, get+1 or-1, α (n) calculates through psychoacoustic model, is used for control and embeds energy, makes it not influence sense of hearing.If Y (n) is the frequency domain amplitude of y (n), then

Y(n)＝X(n)·10 ^{w(n)·S·α(n)}

(3)

Combine original phase information to do the IFFT conversion again the signal Y (n) of embed watermark, accomplish the embedding of a frame.Do overlap-add again between each frame that the completion watermark embeds, recover time-domain signal, obtain the audio frequency of water mark inlaying, accomplish the telescopiny of watermark, the process that watermark embeds is seen Fig. 3.

5. the application of psychoacoustic model

This algorithm utilizes ISO/IEC 11172-3 ^[5]The psychoacoustic model of definition calculates and shelters model, and the details of relevant calculation is at document ^{[5] [6]}Detailed description is all arranged.In the MPEG compression process, to-be-processed frequency domain is divided into some subbands, and psychoacoustic model calculates the letter of each subband and covers than (signal to mask ratios SMR), reflects listened to the degree in audio frequency of this subband.Calculating good SMR is that Bit Allocation in Discrete provides foundation.For the higher subband of SMR, will distribute to more bits, can make this subband that littler quantizing noise is arranged like this.That is to say that in the higher place of SMR, the signal of this subband accounts for leading in the sense of hearing, so will reduce change degree as far as possible to this subband.In carrying out the process that watermark embeds, the energy that a certain frequency embeds is regulated according to SMR, make the lower place of SMR embed more watermark, also be about to watermark and more be not embedded into and go in the part of discovering.Like this, the watermark of embedding just can dynamically be adjusted according to the characteristic of audio frequency itself, makes the whole subjective distortion reduction that under the situation that as far as possible increases the watermark energy that embeds, audio frequency is caused.

2, watermarking algorithm decoding principle

1. relevant decoding

According to embedding formula (3), the frequency number of establishing embed watermark in the frame data is N, then to the signal Y of k frame embed watermark ^k(n) being relevant result with spread spectrum code sequence w (n) is:

$C^k=\underset{n}{\mathrm{\&Sigma;}}w\left(n\right)\&CenterDot;{\log }_{10}\left(Y^k\left(n\right)\right)$

$=\underset{n}{\mathrm{\&Sigma;}}w\left(n\right)\&CenterDot;({\log }_{10}{X}^k\left(n\right)+{\mathrm{\&alpha;}}^k\left(n\right)w\left(n\right)S^k)$

$=S^k\underset{n}{\mathrm{\&Sigma;}}{w}^2\left(n\right)\&CenterDot;{\mathrm{\&alpha;}}^k\left(n\right)+\underset{n}{\mathrm{\&Sigma;}}w\left(n\right)\&CenterDot;x^k\left(n\right)---\left(4\right)$

Result's first is the auto-correlation item of w (n), the positive and negative S that depends on ^kValue, second portion is the simple crosscorrelation item, both are uncorrelated in theory, thereby should be much smaller than the auto-correlation item.

Because watermark is embedded in the dB territory of frequency-region signal, the watermark signal of embedding will be irrelevant with the amplitude of signal, thereby make watermarking algorithm can resist the operations such as amplitude adjustment to time-domain signal.If Y ^k(n) for embedding signal Y ^k(n) through convergent-divergent v signal doubly, then:

$C^{k\&prime;}=\underset{n}{\mathrm{\&Sigma;}}w\left(n\right)\&CenterDot;{\log }_{10}\left(Y^{k\&prime;}\left(n\right)\right)$

$=\underset{n}{\mathrm{\&Sigma;}}w\left(n\right)\&CenterDot;\left({\log }_{10}v{Y}^k\left(n\right)\right)$ (5)

$=\underset{n}{\mathrm{\&Sigma;}}w\left(n\right)\&CenterDot;({\log }_{10}{Y}^k\left(n\right)+{\log }_{10}v)$

$=C^k+{\log }_{10}v\&CenterDot;\underset{n}{\mathrm{\&Sigma;}}w\left(n\right)$

During as

, the amplitude of relevant score and original signal is irrelevant.So when choosing spreading code, need choose the spreading code of balance, promptly spreading code+1 ,-1 number equates.

Also adopted the structure of balance modulation when watermark embeds, the symbol that preceding K frame embeds is reverse with back K frame, calculates the result of this 2K frame respectively, and confirms final bit information based on following formula.

$I=\underset{k=0}{\overset{K-1}{\mathrm{\&Sigma;}}}{C}^k-\underset{k=K}{\overset{2K-1}{\mathrm{\&Sigma;}}}{C}^k---\left(6\right)$

When I＞0, decoded result is 1, and when I＜0, decoded result is 0.

With formula (4) substitution (6),

$I=\underset{k=0}{\overset{K-1}{\mathrm{\&Sigma;}}}{C}^k-\underset{k=K}{\overset{2K-1}{\mathrm{\&Sigma;}}}{C}^k$

$=S\&CenterDot;\underset{n}{\mathrm{\&Sigma;}}\underset{k}{\mathrm{\&Sigma;}}{\mathrm{\&alpha;}}^k\left(n\right)+\underset{n}{\mathrm{\&Sigma;}}w\left(n\right)\&CenterDot;(\underset{k=0}{\overset{K-1}{\mathrm{\&Sigma;}}}{x}^k\left(n\right)-\underset{k=K}{\overset{2K-1}{\mathrm{\&Sigma;}}}{x}^k\left(n\right))$

$=S\&CenterDot;\underset{n}{\mathrm{\&Sigma;}}\underset{k}{\mathrm{\&Sigma;}}{\mathrm{\&alpha;}}^k\left(n\right)+\underset{n}{\mathrm{\&Sigma;}}w\left(n\right)\&CenterDot;x^{\&prime;}\left(n\right)---\left(7\right)$

Compare with formula (4), the x in the simple crosscorrelation item (n) becomes

and is made as x ' (n).Because signal is steady in short-term; X (n) between adjacent a few frame has correlativity; When

was relevant with

, x ' variance (n) was less than x (n).If variable x ₁(n), x ₂(n) for obeying the stochastic variable of identical Gaussian distribution, i.e. x ₁(n): N (0, σ _x), x ₂(n): N (0, σ _x), then the variance of its difference is:

$E{(x_1-x_2)}^2=E{\left(x_1\right)}^2+E{\left(x_2\right)}^2-2E(x_1\&CenterDot;x_2)$

$={\mathrm{\&sigma;}}_x^2+{\mathrm{\&sigma;}}_x^2-2{\mathrm{\&rho;}}_{x_1{x}_2}\&CenterDot;{\mathrm{\&sigma;}}_x\&CenterDot;{\mathrm{\&sigma;}}_x$

$=2(1-{\mathrm{\&rho;}}_{x_1{x}_2}){\mathrm{\&sigma;}}_x^2---\left(8\right)$

Work as x ₁(n), x ₂It is (n) relevant,

Approach 1, so according to above analysis, x ' variance (n) will be much smaller than x (n).According to Kirovski ^[2]The analysis that provides in the literary composition, the false-alarm P of spread-spectrum watermark system _FAWith false dismissal P _MDProbability is respectively

$P_{\mathrm{FA}}=\frac{1}{2}\mathrm{erfc}\left(\frac{\mathrm{\&tau;}\sqrt{N}}{{\mathrm{\&sigma;}}_x\sqrt{2}}\right)---9)$

$P_{\mathrm{MD}}=\frac{1}{2}\mathrm{erfc}\left(\frac{\left(E\right[x\&CenterDot;w]-\mathrm{\&tau;})\sqrt{N}}{{\mathrm{\&sigma;}}_x\sqrt{2}}\right)---\left(10\right)$

Because complementary error Function e rfc (x) is a subtraction function, the minimizing of variance item makes error probability reduce in the formula (9) (10).Like this, utilizing in short-term stably, correlativity can make the decoding error rate descend greatly.

2. synchronizing process:

Synchronizing process was divided into for two steps, i.e. frame synchronization and data sync.The process of frame synchronization is exactly to confirm the process of every frame decoding position.We confirm decoded positions through the method that in watermark signal, adds synchronizing signal in advance.Synchronizing signal is positioned at the front of watermark information segmentation, and adopts different spreading codes with information bit, because each synchronization bit has all adopted different spreading codes to be used to identify different segmentations; So, do not knowing that synchronization bit belongs under the situation of which segmentation, need be correlated with according to the possibility of all spreading codes; So the selection of synchronization bit figure place has determined to carry out how many times and has been correlated with, and is too much like figure place; Associated numbers of times also increases, and figure place is very few, then can make the segmentation lengthening; Influence time resolution should be selected during use as required.Doing when relevant, get the maximum spreading code sequence number of relevant score and be the corresponding spreading code sequence number of this synchronization bit, thereby confirm the segmentation positional information.

For guaranteeing that different spreading codes can correctly identify its place segmentation, require the simple crosscorrelation score between the different spreading codes as far as possible little.Be between the different spreading codes quadrature or relevant score less than certain level.The spreading code that native system is chosen is a balance Gold code ^[7], have good their cross correlation between same group the Gold sign indicating number.Simultaneously, the Gold sign indicating number+1 ,-1 number only differs one, satisfies native system to spreading code desired equilibrium response simultaneously.

In the process of search synchronization bit, we adopt the relevant method of sliding, and set a sliding step and are correlated with.The position of maximum of getting relevant score then is as the position of frame synchronization.Finding suitable frame lock bit postpone, also to carry out data sync, the reference position of specified data.At first we will check that the decoded result of synchronization bit is whether consistent with the synchronizing information that embeds, as inconsistent, and the synchronization failure of this segmentation then.Before beginning synchronously, need to find just can carry out the decoding of information bit a complete information cycle, otherwise information code to be imperfect, decoded result is meaningless.Be introduced into synchronous mode 1 this moment, this pattern requires interior each segmentation of one-period all to meet the frame synchronization condition, and the continuity between the segmentation meets the rule that is provided with in advance.After finding one-period, get into synchronous mode 2.If a segmentation of next finding meets following periodicity condition simultaneously, then think synchronous: the spreading code sequence number that 1) obtains and synchronization bit information and previous cycle consistent 2) establish frame lock bit and be changed to Pos ₂Be changed to Pos with the frame lock bit of a last segmentation ₁, the length on segments theory is L _f, promptly satisfy condition | Pos ₂-Pos ₁-L _f|＜L _Sync, L _SyncConfirm according to experiment.Also not synchronously, establishing this moment decoded positions is Pos after the distance of a segmentation has been passed through in decoding ₃, promptly satisfy condition | Pos ₃-Pos ₁-L _f|＞L _m, L _mAlso confirm according to experiment.Then withdraw from synchronous mode 2, lock-out is searched for synchronous mode 1 state again.

The process flow diagram of synchronous mode is as follows, after going up synchronously, can solve the information code of this segmentation.

3. utilize correlation properties to reduce mistake synchronously:

Sliding when relevant, the curve map of relevant score should present the characteristic of spike, and is as shown in Figure 5, and this characteristic can effectively be used for reducing mistake synchronously, avoids false-alarm, and promptly certain should synchronization position the situation that the decoding score is crossed threshold value once in a while not occur.Because the spreading code that synchronization bit and information bit adopt is inconsistent, so, in the time of synchronously; Information bit can be regarded the situation that does not have embed watermark as, because the structure of balance modulation, decoding is when separating the balance modulation; The signal of the N frame signal of water mark inlaying and a preceding N frame of bit does not have bigger score yet when subtracting each other before, and this score and bit decoding score are reverse; In like manner, behind a bit signal of N frame not with the N frame signal of water mark inlaying when subtracting each other, also have bigger score.And do not have this characteristic around the peak value that part of water mark inlaying does not occur at random.The mean distance that statistics obtains forward peak and reverse peak is L _p, then with frame synchronization position ± L _pThe decoding score C (k) of scope multiply by following window function W (k)

$W\left(k\right)=\left\{\begin{array}{c}1,\left(\right|k-k_{\mathrm{sync}}|\&le;L_p/2)\\ -1,(L_p/2<|k-k_{\mathrm{sync}}|\&le;L_p)\end{array}\right.---\left(11\right)$

K wherein _SyncBe the sync bit point, obtain a mark after adding up

$S_w=\underset{k}{\mathrm{\&Sigma;}}C\left(k\right)\&CenterDot;W\left(k\right)---\left(12\right)$

When judging synchronously, this condition is considered to come in, promptly require score S _wSurpass a certain threshold value, can effectively reduce mistake synchronously.

4. segment information splicing:

After also decoding synchronously, can obtain a lot of segmentation decoding results and its positional information.Finally, we can obtain information service time of audio frequency through these segmentation decodings result.We need be stitched together some segmentations that occur continuously and meet periodic regularity.Judge to splice and to satisfy following condition simultaneously: 1) frame synchronization position Pos ₂Frame synchronization position Pos with a last segmentation ₁Meet | Pos ₂-Pos ₁-L _f|＜L _Sync, L _fBe the length on the segments theory, L _SyncConfirm according to experiment.2) the spreading code sequence number that obtains for before the spreading code sequence number next one sequence number 3 that should occur) decoded result coming to the same thing of previous periodic segment with it.

For some isolated segmentation that occurs, promptly can't directly be spliced to a segmentation in the complete cycle through above condition, when meeting following condition, also be spliced to together.1) frame synchronization position Pos ₂Frame synchronization position Pos with a last segmentation ₁Meet | Pos ₂-Pos ₁-L _sL _f|＜L _Sync, L _f, L _sWith L _SyncDefine identical with before.2) the spreading code sequence number that obtains for before L after the spreading code sequence number _sThe individual sequence number that should occur 3) decoded result coming to the same thing of previous periodic segment with it.

First is fragmented into the sampled point distance of last segmentation to calculate the good segmentation of splicing.Then, obtain temporal information through sampling rate.

5. watermark information reduction:

If the decoded result sequence of the segmentation that stitchs together is { e _i'.Earlier this result is carried out the BCH decoding and obtain information { c _i'.Subsequently, again with watermark information { c _i' decipher through RSA Algorithm.Earlier with { c _i' be expressed as integer c '.Then, according to the deciphering formula of RSA Algorithm

c′ ^d＝m′(mod?M)

(13)

Wherein d is a key, and the key that calculates when encoding before is identical.

Then with m be expressed as binary sequence m ' _i.The decrypting process of completion information.

It should be noted last that above embodiment is only unrestricted in order to technical scheme of the present invention to be described.Although the present invention is specified with reference to embodiment; Those of ordinary skill in the art is to be understood that; Technical scheme of the present invention is made amendment or is equal to replacement, do not break away from the spirit and the scope of technical scheme of the present invention, it all should be encompassed in the middle of the claim scope of the present invention.

Claims (7)

1. method of carrying out copyright management with digital audio frequency watermark, this method comprises the extraction detection method of the embedding grammar and the watermark information of watermark information, wherein,

The embedding grammar of described watermark information comprises following steps:

(1-1) watermark information pre-service is carried out Bose-Chaudhuri-Hocquenghem Code after at first watermark information being encrypted again, obtains watermark encoder information sequence { e _i, 0≤i＜L _B, L wherein _BLength for the watermark information coding;

(1-2) watermark encoder information { e to obtaining _i, 0≤i＜L _BCarry out framing encapsulation;

At first, the watermark encoder information that the matchmaker supports copyright management information that contains that embeds is divided into L _sSection, and with the different different segmentations of spreading code sign; Then, all add a bit synchronization position before the segmentation with each watermark encoder information, information bit adopts different spreading codes with synchronization bit, obtains watermark and embeds information { w _i, 0≤i＜L _S+ L _B;

(1-3) watermark that obtains is embedded information { w _i, 0≤i＜L _S+ L _BSequence carries out base band balance modulation, the watermark with a bit during modulation embeds Information hiding among the sound signal of continuous several frames, obtains the signal S after the balance baseband modulation;

(1-4) signal S is modulated in the original audio signal, the modulation formula is following;

Y(n)＝X(n)·10 ^{w(n)·s·α(n)}

Wherein, { X (n), 0＜n＜N} treat the amplitude of the sound signal of embed watermark at frequency domain, and N is the length of the spreading code of use, and X (n) unit is dB; { w (n), 0＜n＜N} are spreading code, and this spreading code is that information bit or synchronization bit are confirmed by the type that embeds bit, and α (n) calculates through psychoacoustic model, are used for control and embed energy;

(1-5) combine original phase information to do the IFFT conversion again the signal Y (n) of embed watermark at last; Accomplish the embedding of a frame, will accomplish again between each frame that watermark embeds and do overlap-add, recover time-domain signal; Obtain the sound signal of embed watermark, thereby accomplish the telescopiny of watermark;

The extraction detection method of described watermark information comprises following steps:

(2-1) relevant decoding, the process of this process and coding is opposite, obtains false-alarm probability and false dismissal probability through relevant decoding;

(2-2) synchronizing process, this synchronizing process comprises frame synchronization and data sync, is used to solve the information code of synchronous segmenting;

(2-3) segment information splicing, the segmentation that will occur and meet some completion decodings of periodic regularity according to the splicing decision condition continuously is stitched together, and obtains information service time of audio frequency;

Wherein, the splicing decision condition is:

1) frame synchronization position Pos ₂Frame synchronization position Pos with a last segmentation ₁Meet | Pos ₂-Pos ₁-L _f|＜L _Sync, L _fBe the length on the segments theory, L _SyncConfirm according to experiment;

2) the spreading code sequence number that obtains for before the spreading code sequence number next one sequence number that should occur;

3) decoded result coming to the same thing of previous periodic segment with it;

Satisfy above three conditions simultaneously and just splicing implemented in segmentation, and calculate splicing first be fragmented into the sampled point distance of last segmentation, obtain the service time of audio frequency then through sampling rate;

(2-4) watermark information reduction, the extraction of accomplishing watermark information detects.

2. method of carrying out copyright management with digital audio frequency watermark according to claim 1 is characterized in that, described synchronizing process adopts correlation properties to reduce mistake synchronously, and concrete steps are:

Limit score S _wSurpass a certain threshold value, this CALCULATION OF PARAMETERS formula is following:

$S_w=\underset{k}{\mathrm{\&Sigma;}}C\left(k\right)\&CenterDot;W\left(k\right)$

Wherein, C (k) is the decoding score;

W (k) is a window function, and computing formula is following:

$W\left(k\right)=\left\{\begin{array}{c}1,\left(\right|k-k_{\mathrm{sync}}|\&le;L_p/2)\\ -1,(L_p/2<|k-k_{\mathrm{sync}}|\&le;L_p)\end{array}\right.$

Wherein, L _pThe mean distance at the forward peak that obtains for statistics and reverse peak, k _SyncBe the sync bit point.

3. method of carrying out copyright management with digital audio frequency watermark according to claim 1 is characterized in that, described watermark information pre-service concrete steps are:

At first utilize several rules to produce the PKI e and the key d of RSA Algorithm;

The encryption formula that utilizes RSA Algorithm then is to the watermark information sequential encryption;

Information { c after the encryption _iThrough Bose-Chaudhuri-Hocquenghem Code, form the watermark encoder information { e that encrypts _i, 0≤i＜L _B.

4. method of carrying out copyright management with digital audio frequency watermark according to claim 1 is characterized in that, described synchronizing process adopts the relevant method of sliding in the process of search synchronization bit, and step is:

At first, setting a sliding step is correlated with; The position of maximum of then, getting relevant score is as the position of frame synchronization; At last, find suitable frame lock bit postpone, carry out data sync, the reference position of specified data.

5. method of carrying out copyright management with digital audio frequency watermark according to claim 1 is characterized in that, described watermark information reduction comprises following steps:

The decoded result sequence of at first, establishing the segmentation that stitchs together is { e _i', earlier this result is carried out the BCH decoding and obtain information { c _i'; Then, again with watermark information { c _i' decipher through RSA Algorithm, obtain m '; At last with m be expressed as binary sequence m ' _i, the decrypting process of completion information.

6. method of carrying out copyright management with digital audio frequency watermark according to claim 1 is characterized in that, described spreading code is a balance Gold code.

7. method of carrying out copyright management with digital audio frequency watermark according to claim 1 is characterized in that, when described segment information splices for some isolated concrete treatment step of segmentation that occurs is:

, segmentation is spliced to together when meeting following condition simultaneously:

1) frame synchronization position Pos ₂Frame synchronization position Pos with a last segmentation ₁Meet | Pos ₂-Pos ₁-L _sL _f|＜L _Sync

2) the spreading code sequence number that obtains for before L after the spreading code sequence number _sThe individual sequence number that should occur;

3) decoded result coming to the same thing of previous periodic segment with it.

Images